Multi-Vector Search

Zvec supports multi-vector queries, allowing you to combine different embeddings in a single search.

When querying multiple vector embeddings, Zvec retrieves candidates from each vector space independently and then fuses them into one relevance-ordered list. Because scores from different vector spaces might not be directly comparable, you must choose a suitable re-ranking strategy.

Prerequisites

This guide assumes:

• You have opened a collection with multiple vector fields
• You're familiar with the basic vector querying concepts. If not, please review the single-vector search guide

Performing Multi-Vector Search

In Python, pass a list of query specifications to query() and provide a fusion strategy through reranker. In Node.js, pass the sub-queries to multiQuerySync() or multiQuery() and configure rerank.

This example queries both dense_embedding and sparse_embedding and uses WeightedReRanker to combine their results. Weights are positional and must follow the same order as queries.

import zvec

result = collection.query(  
    topk=3,  # Number of final documents returned after re-ranking
    queries=[  # List of query specifications — one for each embedding space to search
        zvec.Query(field_name="dense_embedding", vector=[0.1] * 768),           
        zvec.Query(field_name="sparse_embedding", vector={1: 0.1, 37: 0.43}),   
    ],
    reranker=zvec.WeightedReRanker(  
        weights=[1.2, 1.0],  # dense_embedding, then sparse_embedding
    ),
)
print(result)

Re-ranking Strategies

Zvec provides different re-ranking strategies to combine scores from multiple vector fields.